Download Interactions between attention, context and learning in primary

Vision Research 40 (2000) 1217 – 1226
www.elsevier.com/locate/visres
Interactions between attention, context and learning in primary
visual cortex
Charles Gilbert a,*, Minami Ito b, Mitesh Kapadia a, Gerald Westheimer a
b
a
The Rockefeller Uni6ersity, 1230 York A6enue, New York, NY 10021, USA
Laboratory for Neural Control, National Institute for Physiological Sciences, Myodaiji, Okazaki 444 -8585, Japan
Received 4 June 1999; received in revised form 12 October 1999
Abstract
Attention in early visual processing engages the higher order, context dependent properties of neurons. Even at the earliest
stages of visual cortical processing neurons play a role in intermediate level vision — contour integration and surface
segmentation. The contextual influences mediating this process may be derived from long range connections within primary visual
cortex (V1). These influences are subject to perceptual learning, and are strongly modulated by visuospatial attention, which is
itself a learning dependent process. The attentional influences may involve interactions between feedback and horizontal
connections in V1. V1 is therefore a dynamic and active processor, subject to top-down influences. © 2000 Elsevier Science Ltd.
All rights reserved.
Keywords: Primary visual cortex; Perceptual learning; Contextual influences; Receptive fields; Visuospatial attention
To explore the role of attention in any cortical area
one has to take into account the role that area plays in
the processing of visual information. The primary visual cortex (V1) is engaged in much more complex
processes than previously believed, and the very concept of the receptive field has undergone a substantial
modification. The processes included under the rubric
of intermediate level vision, such as contour integration
and surface segmentation, may be represented at the
earliest stages of the visual cortex pathway. This view is
in sharp contrast to the classical notion about V1
analyzing only the most rudimentary attributes of the
visual world, such as local orientation or the position in
depth of local features. Instead, the responses of neurons in V1 are dependent on the precise geometric
relationships between line segments or texture elements
within the receptive field and contours and surfaces
extended for considerable distances outside the receptive field. The response properties of cells to features in
the visual scene are highly dependent on the context
within which those features are placed, and those con* Corresponding author. Tel.: +1-212-3277670; fax: + 1-2123277844.
E-mail address: [email protected] (C. Gilbert)
textual influences are the properties most modulated by
attention. In addition, the contextual modulation is
influenced by learning, and the attentional influences
themselves are subject to learning effects.
The idea that the response of neurons to local attributes is dependent on the global characteristics of
contours and surfaces is consonant with the basic tenet
of Gestalt psychology. The principle, as put forward by
Wertheimer, holds that the ‘‘behavior of the whole
cannot be derived from its individual elements nor from
the way these elements fit together; rather the opposite
is true: the properties of any of the parts are determined
by the intrinsic structural laws of the whole’’
(Wertheimer, 1923). The contextual modulation of receptive fields, though currently put under the category
of the ‘non-classical’ receptive field, is actually close to
one of the original descriptions of the receptive field by
Stephen Kuffler. He emphasize that ‘‘… not only the
areas from which responses can actually be set up by
retinal illumination may be included in a definition of
the receptive field but also all areas which show a
functional connection, by an inhibitory or excitatory
effect on a ganglion cell. This may well involve areas
which are somewhat remote from a ganglion cell and
by themselves do not set up discharges’’ (Kuffler, 1953).
0042-6989/00/$ - see front matter © 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 2 - 6 9 8 9 ( 9 9 ) 0 0 2 3 4 - 5
1218
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
One now can extend the concept of receptive field even
further to include the potential for dynamic changes on
a range of time scales (from seconds to weeks) and
depending on experience, behavioral task and attention
and expectation.
A number of studies have shown that cells in primary
visual cortex are sensitive to stimuli lying outside their
receptive fields. The classical measure of the receptive
field, known as the minimum response field, is obtained
by placing an optimally oriented line in different positions within the visual field. Complex stimuli, composed
of multiple line segments presented simultaneously,
show strong modulation of cells’ responses relative to
the response to a single line within the receptive field.
This has led to the distinction between the ‘classical’
and ‘non-classical’ receptive field, although even those
properties termed non-classical would fall under
Kuffler’s classic description of the receptive field. The
non-classical properties have been implicated in a range
of roles including contour integration, perceptual fill-in,
surface segmentation and orientation contrast (Blakemore & Tobin, 1972; Maffei & Fiorentini, 1976; Nelson
& Frost, 1978; Allman, Miezin & McGuinness, 1985;
Nothdurft & Li, 1985; Tanaka, Hikosaka, Saito, Yukie,
Fukada & Iwai, 1986; Orban, Guylas & Vogels, 1987;
Gilbert & Wiesel, 1990; Knierim & Van Essen, 1992; Li
& Li, 1994; Kapadia, Ito, Gilbert & Westheimer, 1995;
Lamme, 1995; Sillito, Grieve, Jones, Cuderio & Davis,
1995; Rossi, Rittenhouse & Paradiso, 1996; Zipser,
Lamme & Schiller, 1996; Kastner, Nothdurft & Pigarev, 1997; Levitt & Lund, 1997; Kapadia, Westheimer & Gilbert, 1998, 1999).
An important source of contextual influences is a
plexus of long range horizontal connections that link
cells with widely separated receptive fields. This is a
network of connectivity formed by the axons of cortical
pyramidal cells. Pyramidal and spiny stellate cells represent 80% of the neurons in the cortex, each cell can
form lateral connections of varying extents. The longest
connections run for 5 – 6 mm, from one end of the
axonal arbor to another (Gilbert & Wiesel, 1979, 1983,
1989; Rockland & Lund, 1982; Gilbert, 1992; Bosking,
Zhang, Schofield & Fitzpatrick, 1997). From the point
of view of the neurons receiving these connections, the
area of visual space represented by the area of cortex
from which they integrate input can be an order of
magnitude larger than their own receptive fields. The
horizontal connections are distributed in discrete clusters of axon collaterals, showing an underlying specificity in the connections. This specificity is related to the
columnar functional architecture of cortex, with connections linking columns of similar orientation preference and cells with widely separated receptive fields
(Gilbert & Wiesel, 1979, 1983, 1989; Malach, Amir,
Harel & Grinvald, 1993; Weliky, Kandler, Fitzpatrick
& Katz, 1995; Bosking et al., 1997; Kisvarday, Toth,
Rausch & Eysel, 1997). The registration between the
horizontal connections and the columnar functional
architecture mirrors the perceptual phenomenon of
greater saliency for contours made of similarly oriented
line segments, and is reflected in the geometry of contextual influences (Wertheimer, 1923; Grossberg &
Mingolla, 1985; Ullman, 1990; Field, Hayes & Hess,
1993; Wolfson & Landy, 1999).
One must also allow for possible contributions from
feedback connections in the global response properties
of cortical neurons. It has been suggested, by looking at
the effect of inactivation of V2 on center-surround
interactions in V1 (Hupe, James, Payne, Lomber,
Girard & Bullier, 1998) and because of the delayed time
course of responses to texture discontinuities (Lamme,
1995; Zipser et al., 1996) that at least some of the
contextual influences are derived from feedback connections from higher order cortical areas to area V1. As
shown below, however, the contextual facilitation we
observe follows the full time course of the response. In
any event, it is not clear that routing information
through V2 back to V1 requires a substantial additional
delay, and there might be equivalent delays in the
lateral influences originating from V1. The spatial characteristics and columnar specificity of horizontal connections and the rules of perceptual interactions
(contour saliency and contrast enhancement) are quite
comparable. The interplay between attention and contextual influences, as we will develop, does argue for an
interaction between intrinsic horizontal connections
and feedback connections in the final outcome of visuospatial integration.
The contextual influences for cells in V1 show a
powerful facilitation that depends on the precise relative placement of line segments inside and outside the
receptive field. Psychophysical experiments measuring
the detectability of a line segment show a similar facilitation when the line is associated with a nearby,
collinear line segment (Dresp, 1993; Polat & Sagi, 1993,
1994; Kapadia et al., 1995). Subjects can detect a line at
40% lower contrast when paired with a collinear line
than when presented in isolation. The response of cells
in the superficial layers of the primary visual cortex of
alert monkeys to a single line inside the receptive field
can be facilitated as much as 3-fold by adding a line
segment outside the receptive field (Fig. 1, see Kapadia
et al., 1995; Polat, Mizobi, Pettet, Kasamatsu & Norcia, 1998). The facilitation is maximal when the second
line is collinear, adjacent to and of the same orientation
as the first. The fall off in facilitation in contrast
detection thresholds and the facilitation in the responses of cells to contextual stimuli, when measured at
the same eccentricity, follow the same dependency on
the separation between the lines along the collinear
axis, on the lateral offset between the lines, and on the
relative orientation of the two lines. The spatial scale of
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
these effects also mirror the scale of the long range
horizontal connections. The fact that a line outside the
receptive field by itself does not activate the cell, yet
when presented in conjunction with a line within the
receptive field can triple the response of the cell, indicates a substantial degree of non-linearity in cell responses. The facilitation can be blocked by placing a
perpendicular line between the two collinear lines, suggesting a role of the facilitation in linkage of line
segments that compose a contour.
The contextual modulation of cell responses demonstrates that one cannot predict the response of a cell to
a complex visual scene from its responses to the individual elements of the scene presented in isolation. This
non-linearity can be accounted for by the nature of the
e.p.s.p. generated by the horizontal connections, which
is much larger when the cell is depolarized than when it
Fig. 1. Facilitation in detection and responses to a target line by a
collinear flanking line. (A) Subjects were asked to report the presence
of a line presented in the near periphery (at 4° eccentricity) at a range
of luminances, when presented in isolation and in conjunction with a
flanking line. As measured by the psychometric curves, when the
target was flanked by a colinear line, it could be detected at much
lower contrast than when it was presented alone. This effect was
diminished as the flanking line was separated from the target, shifted
along an axis orthogonal to its orientation, or changed in relative
orientation. (B) Cells were recorded in the superficial layers of area
V1. One line was placed within the receptive field and a second line
was placed outside the receptive field. The cell did not respond to the
second line presented by itself, but its response was increased 2- or
3-fold when the second line was presented in conjunction with the line
within the receptive field. The facilitation dropped off with increasing
separation along the collinear axis, by a lateral offset between the two
lines, and by changing the orientation of the flanking line. At
matched eccentricities the contrast reduction by a flanking line measured psychophysically and the facilitation measured physiologically
showed dependency on line separation and relative orientation over
similar spatial scales (Kapadia et al., 1995).
1219
Fig. 2. Averaged response histograms among cells showing contextual
facilitation The response to contextual stimuli at the relative angle
and attentional state showing the highest facilitation was selected for
each cell. The response of each cell was normalized to the same unit
area. The facilitation extended throughout the entire time course of
the cells’ responses. The black underlining indicates stimulus presentation time. The gray bar indicates the time period of significant
facilitation (P B0.05). From Ito and Gilbert (1999).
is at rest (Hirsch & Gilbert, 1991). As a result, the
horizontal input, activated by a contextual stimulus, is
stronger when the cell is simultaneously activated by
interlaminar connections, which would be activated by
stimuli lying within the receptive field.
In the presence of more complex visual environments, and under distributed attention (see below) the
facilitation is seen not just with stimuli presented at
threshold contrast but with suprathreshold stimuli as
well (Ito & Gilbert, 1999). The time course of the
facilitation follows that of the response itself, and the
effect of attention, which we will outline below, is
exerted on the earliest components of the response.
Over a recorded population of cells in the superficial
layers of primary visual cortex, the presence of a contextual line produces an average 2-fold facilitation, and
this extends throughout the full time-course of the
response (Fig. 2). In this series of experiments, the data
were selected from the animal’s attentional state that
showed the maximal amount of facilitation for each
cell.
The idea that the facilitation relates to the integration
and saliency of contours is supported by experiments in
which a contour is embedded in a complex background
(Fig. 3). Here, a line was placed within the receptive
field at the optimum orientation for the cell, and this
line was surrounded by an array of randomly positioned and oriented lines. Individual line segments in
the surrounding array were then moved into alignment
with the line within the receptive field, forming a long
straight line extending well beyond the receptive field.
When there is a single line within the receptive field
surrounded by the complex background, the response
of the cell is greatly inhibited by the surround. As lines
1220
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
are shifted into alignment with the line within the
receptive field, the cell is lifted from the inhibition. As
the line is further extended by adding line segments, the
response can be further facilitated to levels much
greater than those seen when the line within the receptive field is presented in isolation. This shows that the
responses of neurons within V1 are as dependent on the
global characteristics of contours extending well beyond the boundary of the classical receptive field as
they are upon the attributes of features lying within the
receptive field. It also indicates that the process of
intermediate level vision may find its substrate as early
in the visual cortical pathway as the primary visual
cortex (Kapadia et al., 1995; Polat et al., 1998). The
evidence for V1 participation in contour integration, in
the sense of contour saliency, extends the evidence for
responses in V2 to illusory contours (von der Heydt
and Peterhans, 1989).
With this as background, one can then pursue the
question of the role of attention in modulating the
responses of cells in V1. Since the responses of cells in
V1 depend on the precise geometric juxtaposition of
lines inside and outside the receptive field, it is important to take the specificity of their responses to stimuli
Fig. 3. Facilitation in responses by collinear stimuli in a complex background. Top, the range of stimuli used, from a single bar within the
receptive field (1), one bar inside and one or two flanking bars (2, 3), a single flanking bar alone (4), the inside bar embedded in a complex
background of randomly positioned and oriented bars (5), within that background adding one, two and 4 collinear flanking bars (6, 7, 8). Bottom,
the cells show varying amounts of facilitation to the flanking bars, but this facilitation is greater in the presence of the complex background. The
background inhibits the response of the cell, but the presence of a global contour extending beyond the receptive field removes the inhibition and
produces responses greater than that obtained by the isolated bar within the receptive field. From Kapadia et al. (1995).
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
Fig. 4. Stimulus array used to explore the role of visuospatial
attention on the contextual facilitation in perceived brightness and on
the responses of cells in primary visual cortex. Four sets of target and
flanking lines were presented symmetrically around the fixation point
(the target stimuli, to which the subject had to respond, were the
inner ones of the pairs). A reference line was presented near the
fixation spot, and the subject had to report whether one of the target
lines was brighter or dimmer than the reference. On some trials the
subject was cued to the position of the odd target prior to the
presentation of the stimulus array (focal attention) and on others was
cued to all four positions (distributed attention). For the physiological experiments, the focal attention trials were further divided into
trials in which attention was focussed on the receptive field position
and trials in which attention was focussed onto a position away from
the receptive field. The orientation and positions of the bars were
adjusted such that the target bar was centered within the receptive
field and oriented at the optimum orientation of the cell, with the
remaining bars distributing symmetrically around the fixation point.
From Ito, Westheimer and Gilbert (1998) and Ito and Gilbert (1999).
Fig. 5. Effect of attention on contrast discrimination threshold and
contextual facilitation. Top, the stimulus array was either a set of
four target lines and a reference line near the fixation spot, or four
pairs of target and flanking lines. Bottom, there were marked differences in both the threshold and facilitation between distributed
attention and focal attention trials. The most marked difference was
observed in the facilitation in perceived brightness by the flanking
line. From Ito et al. (1998).
of higher order complexity when exploring attentional
effects. We therefore designed a stimulus array and a
behavioral paradigm that would probe the effect of
visuospatial attention on contextual influences as well
as upon single line stimuli within the receptive field.
The experiments were designed to provide a behavioral
measure of these effects as well as to explore the effect
1221
of attention on cell responses. Here we used
suprathreshold stimuli and a discrimination task rather
than the detection task used initially. The subjects, both
human and non-human primate, were asked to report
whether a target line, presented at 3.5° eccentricity, was
brighter or dimmer than a reference line presented
adjacent to the fixation spot. In addition to the target
line, on some trials there was a flanking collinear line to
which the subjects did not attend. Four sets of target
and flanking lines were presented, placed symmetrically
around the fixation point at the 45/135/225/315° meridia (Fig. 4). On some trials, termed the distributed
attention trials, the subject was cued to all four target
positions. In these the subject had to determine whether
any of the four target lines were brighter or dimmer
than the reference line. On other trials, termed the focal
attention trials, the subject was cued to the single
position in which the odd target would be presented. In
this task, even though the target stimuli were presented
at suprathreshold stimuli, there was still a large facilitatory effect of the flanking line. In the presence of the
flanking line, the psychometric curve on the perceived
brightness of the target line relative to the reference line
was shifted to the left. The shift of the mean in the
curve provided a measure of the change in perceived
brightness, and the slope provided a measure of the
discrimination threshold. One can then determine the
relative effects of distributed versus focal attention on
the threshold and on the facilitation.
For naı̈ve subjects, there was a substantial difference
in the perceptual facilitation of a flanking line between
distributed and focal attention (Fig. 5). The effect was
much stronger under distributed attention. In effect, if
one is already attending to a stimulus, there is little
additional saliency derived from the stimulus characteristics themselves. There was also an influence of attention on discrimination threshold, consistent with earlier
studies (Beck & Ambler, 1973; Cohn & Lasley, 1974;
LaBerge, 1983; Eriksen & St. James, 1986; Krose &
Julesz, 1989; Lindblom & Westheimer, 1992a,b; Balz &
Hock, 1997). In examining the effect of attention on the
responses of cells, one can further subdivide the category of focal attention into attention to the receptive
field position and attention away from the receptive
field position. An example of a cell recorded in one
monkey, SA, is shown in Fig. 6. Under distributed
attention the cell showed the nearly 3-fold facilitation
by contextual stimuli that we had seen previously.
Under focal attention, the facilitation disappeared.
Because the animals receive different cues under focal
and distributed attention, and because the sizes of
receptive fields in V1 are relatively small, it is natural to
ask whether predictive eye movements might influence
the results. The animals were highly trained on the task
of fixation, and any movement outside of the fixation
window (0.5° radius) aborted the trial. The fixation had
1222
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
to be more precise than that allowed by the window,
since the animals had to detect a slight dimming of the
fixation spot. Measurements of eye position with a
scleral search coil bore out this argument. There was
less than 0.1° difference in eye position between the
different attentional states, which was not significant
and with no systematic relationship between mean eye
position and the cue position (Fig. 7). There was also
no difference in the variance of eye position between
the different attentional states. Shifting eye position on
the order of 0.1° produced no change in the amount of
facilitation.
An important factor in the attentional influences is
effect of learning. One can train subjects not just on the
attributes of a visual stimulus, but even an attentional
task is itself subject to learning. This was seen in the
focal vs. distributed attention task, where the performance observed under distributed attention converged
to that seen under focal attention (Fig. 8). That is, the
performance under distributed attention after training
improved to the level of focal attention before training.
The improvement effectively represents an ability to
perform ‘multi-focal’ attention, by increasing the number of attentional foci maintained in parallel. This
learning might be equivalent to that described for the
transformation from serial to parallel search tasks by
learning (Sireteanu & Rettenbach, 1995). The improvement shows an interesting specificity for the stimulus
configuration used in training. If the number of targets
and flanks are increased from four to eight (e.g. eight
targets presented at the 0/45/90/135/180/225/270/315°
meridia, equidistant from the fixation point), there is a
return to the original level of performance on the
distributed attention task, but only for the new stimulus
positions — the original positions retain the improved
performance. The improvement is not, however, visuotopically specific (it transfers to visual field locations
Fig. 6. Influence of attention on responses in primary visual cortex.
For this cell, recorded in the superficial layers of primary visual
cortex, when the animal distributed attention to all four target
locations there was a severalfold increase in the response to the target
plus flanking line compared to the target alone response. This facilitation disappeared when the animal was cued to focus attention to one
of the target lines. From Ito and Gilbert (1999).
Fig. 7. Measurement of eye position, receptive field size and contextual facilitation. (A) Minimum response field profile and stimulus
placement. In this example, a single bar was placed at half bar length
intervals, providing a measure of receptive field boundary. The target
stimulus was centered over the position of highest response, and a
second stimulus of equal size was placed outside the receptive field to
ensure that it would not activate the cell by itself. (B) Attentional
modulation of responses of cell shown in A. Contextual facilitation
was maximal when the monkey attended to the receptive field position (focal on), and disappeared when the animal distributed his
attention to all four sites or attended to positions away from the
receptive field. Note that the target alone response was the same for
focal attention on and focal attention away, even though the contextual modulation was very different. (C) There was no predictive eye
drift towards the cued location while we collected the responses
shown in B. The mean eye position under distributed attention is
represented as the center of the graph, and the eye positions measured
under the other conditions are shown relative to this point. For each
attentional state, the averaged position and standard deviation are
indicated. The outer circle represents a distance 0.5° from the center.
The arrow indicates the direction toward the receptive field (RF) of
this cell. (D) Eye positions measured in the entire day’s session,
during which we recorded from the cell shown in this Figure. We
divided data into six groups for each cued location. Description was
same as in C. Arrow and mark indicate corresponding direction
toward cued-location. Since the receptive field position was different
for each cell, directions towards the receptive field are rotated to the
same direction (arrow RF) and the other cued locations are shown
relative to the RF direction. (E) The data shown in B (focal on) was
divided into two groups, one half taken from trials where the eye
fixations were closer to the receptive field of the cell (near) and the
other half from trials was the eye was farther from the receptive field
(far). Though the eye position in these two groups differed by 0.11°,
there was no difference in the degree of facilitation. Taken from Ito
and Gilbert (1999), Fig. 3.
other than those in which the training stimuli were
located). If the original four stimulus array is rotated to
the new positions (e.g. from the 45/135/225/315° meridia to the 0/90/180/270° meridia), one sees the improved level of performance in the new positions.
Effectively the subjects are capable of performing a
mental rotation of the stimulus configuration. The specificity for stimulus configuration, rather than for
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
Fig. 8. Effect of training on attentional task. Top, results for seven
subjects, including to Macaque monkeys. The facilitation under distributed attention converged to that measured under focal attention
after several weeks of training. Bottom, in the Macaques, as a result
of overtraining, the endpoint of training showed opposite effects in
the two monkeys. Monkey SA showed greater facilitation under
distributed attention, and monkey UM showed greater facilitation
under focal attention. From Ito et al. (1998) and Ito and Gilbert
(1999).
visuotopic locus, speaks to an interaction between areas
which represent a global representation of the complex
array and V1, where precise location is represented.
1223
Perceptual learning can originate from a combination
of top-down influences, as suggested from its configuration-dependency, and from local, early cortical processing, as as suggested from its visuotopic specificity (Shiu
& Pashler, 1992; Treisman, Veira & Hayes, 1992;
Ahissar & Hochstein, 1993; Sagi & Tanne, 1994;
Gilbert, 1994; Fahle & Morgan, 1996; Crist, Kapadia,
Westheimer & Gilbert, 1997). In addition to the role of
attention in enabling perceptual learning of stimulus
attributes, it is important to emphasize the role of
perceptual learning of attentional tasks. One cannot,
therefore, expect an equivalency between subjects in
correlating attention with physiology — it is important
to measure performance on the same individuals from
which physiological data are obtained (e.g. one must do
psychophysics with the monkeys used for the microelectrode recording studies).
In the non-human primates we studied, there were
further learning dependent changes as a result of overtraining. As a result, one animal, SA, showed more
contextual facilitation under distributed attention and
the other animal, UM, showed greater contextual facilitation under focal attention (Fig. 8). The fact that the
animals were overtrained on the task leads to the
different effects of attention on facilitation than that
observed before recording and to differences relative to
the results obtained from human subjects. This led to
differences in the attentional effects measured in area
V1, as seen over the population of recorded cells (Fig.
9). For subject SA, cells showed more contextual facilitation under distributed attention than under focal
attention. In contrast, there was little difference in the
Fig. 9. Effect of attention over the population of recorded V1 neurons. The data were collected for the two monkeys in which the psychophysical
measurements were made, SA (A,C) and UM (B,D). Two properties were measured: the mean response to the test line alone under the three
attentional states (A,B) and the mean percentage facilitation by a flanking bar (0% indicates no facilitation, 100% indicates a doubling of the
response). For SA, the facilitation was larger under distributed attention than under focal attention, and for UM the relationship was reversed,
with greater facilitation under focal attention. Note that for both animals the facilitation with focal attention away from the receptive field
position was equivalent to that seen under distributed attention, indicating that the modulation by focal attention was restricted to sites near the
attentional focus. The behavioral conditions in the upper graphs follow the same order as those in the lower graphs. From Ito and Gilbert (1999).
1224
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
target-alone response over the three attentional states.
For subject UM, on the other hand, cells showed more
contextual facilitation under focal than under distributed attention. Again, the target alone responses
were unchanged by attention. Another point worth
emphasizing is that for both animals the facilitation
seen under focal attention away from the receptive field
was similar to that seen under distributed attention, so
that the effect of focal attention was restricted to
positions near the receptive field. The attentional effects
on contextual facilitation within primary visual cortex
are therefore pronounced, changing by a factor of two
or more according to attentional state. The effects on
simple visual stimuli, however, are negligible.
The small effect of attention on the target alone
response, compared with the powerful effect of attention on contextual influences, might account for the
discrepancy with respect to earlier reports showing little
effects of attention in V1 (Wurtz & Mohler, 1976;
Haenny & Schiller, 1988; Motter, 1993; Luck, Chelazzi,
Hillyard & Desimone, 1997). However, the size of
attentional effects has been shown to increase with the
number of distractors (Motter, 1993; Vidyasagar, 1998).
This is consistent with ideas that attention is important
when a target must be found in a cluttered field, that
attention represents a competition between stimuli, and
that attending to one location reduces processing of the
local surround (Engel, 1971; Desimone & Duncan,
1995; Bahcall and Kowler, 1999). The earlier studies
failing to activate V1 generally used quite sparse visual
stimulation and simple tasks. Physiological studies involving a curve tracing task (Roelfesma, Lamme &
Spekreijse, 1998) and several fMRI and EEG studies
(Brefczynski & DeYoe, 1999; Gandhi, Heeger & Boynton, 1999; Martinez, Anllo-Vento, Sereno, Frank, Buxton, Dubowitz, et al., 1999; Somers, Dale, Seiffert &
Tootell, 1999) now show substantial attention effects in
V1. Subjects asked to imagine the appearance of an
object in the absence of visual input (visual imagery)
show activation of V1 (Kosslyn, Thompson, Kim,
Rauch & Alpert, 1995). Taken together, these findings
and those in the current work demonstrate the importance of top-down influences in the activation of V1
when subjects are actively engaged in interpreting the
visual environment. The degree of attentional modulation of responses in V1 is also dependent on the behavioral context. The amount of attentional resources
demanded by the task, for example, plays an important
role (Nakayama & Joseph, 1997). Attention is therefore
not an all or none phenomenon, rather it is dependent
upon the requirements of the task and the nature of the
attended stimuli.
Our results show that attentional effects can be quite
large, by a factor of two or more, when one looks at the
proper psychophysical measurement — contextual facilitation. It is clear, however, that the attentional ef-
fects show several dependencies that must be taken into
account when studying the attention effects in any
cortical area. First, it is important to select the correct
visual stimulus, taking into account the higher order
functional properties of cells and the manner in which
they are analyzing the visual scene. Second, it is also
necessary to take into account the appropriate behavioral context, where the attentional task, both in terms
of kind and difficulty, will determine the nature and
extent of modulation exerted in a given cortical area.
Third, the fact that these effects are subject to learning
emphasizes the importance of taking into account, and
measuring, the performance of each individual subject
in the discrimination task that is under attentional
control. Since these findings, as well as others, emphasize the emergence of attentional influences by an interaction between connections intrinsic to V1 and
interactions between multiple cortical areas, it will be
difficult to isolate the origin of these effects. To the
extent that contextual effects, originating in V1, are
gated by feedback connections, as suggested by the
attention effects, one could imagine that inactivating
the source of feedback would also influence the lateral
interactions within V1. It is important to emphasize,
however, that the time course of the contextual facilitation, and that of the attentional modulation of contextual facilitation (when attention is cued before stimulus
presentation), starts from the beginning of the sensory
response. Therefore the output from V1, possibly from
the very first spike, reflects both the contextual and
attentional influences, decreasing the likelihood that
they originate elsewhere.
The contextual influences on cells in primary visual
cortex indicate that they are selective for much more
complex features in the visual environment than previously thought, and that they play a central role in
contour integration and surface segmentation. These
influences depend on the precise geometric relationships
between stimuli lying inside and outside the receptive
field core, and the interactions between stimuli placed
in the extended receptive field and those lying within
the classical receptive field are most subject to attentional modulation. The response to a simple stimulus,
such as a single oriented line segment, is mediated by
feedforward connections such as thalamocortical and
vertical interlaminar connections within V1. The modulation of these responses by the global characteristics of
contours and surface boundaries extending beyond the
receptive field is likely to be mediated at least in part by
the plexus of long range horizontal connections. One
also must consider feedback connections from higher
cortical areas as potentially playing a part in certain
contextual influences. One possible mechanism underlying the attention effects is a gating or modulation of the
synaptic effects of long range horizontal connections by
feedback connection from higher cortical areas. If so,
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
then one must view the contextual interactions as
emerging from an interaction between the various inputs converging onto neurons in the primary visual
cortex.
The fundamental conclusion from these studies is
that V1 is not simply a passive filter subject to feedforward interactions, but is an active processor, dynamically changing its processes according to top down
influences. To see these influences operate, however,
one must take into account the higher order, integrative
properties of cells in V1, the task dependent nature of
these properties, and the role of learning in modulating
these influences.
Acknowledgements
This work was supported by NIH grant EY07968 to
C.D.G.
References
Ahissar, M., & Hochstein, S. (1993). Attentional control of early
perceptual learning. Proceedings of the National Academy of Sciences USA, 90, 5718 –5722.
Allman, J., Miezin, F., & McGuinness, E. (1985). Stimulus specific
responses from beyond the classical receptive field–neurophysiological mechanisms for local global comparisons in visual neurons. Annual Re6iew of Neuroscience, 8, 407–430.
Bahcall, D. O., & Kowler, E. (1999). Attentional interference at small
spatial separations. Vision Research, 39, 71–86.
Balz, G. Y., & Hock, H. S. (1997). The effect of attentional spread on
spatial resolution. Vision Research, 37, 1499–1510.
Beck, J., & Ambler, B. (1973). The effects of concentrated and
distributed attention on peripheral acuity. Perception and Psychophysics, 14, 225 – 230.
Blakemore, C., & Tobin, E. A. (1972). Lateral inhibition between
orientation detectors in the cat’s visual cortex. Experimental Brain
Research, 15, 439 – 440.
Bosking, W. H., Zhang, Y., Schofield, B., & Fitzpatrick, D. (1997).
Orientation selectivity and arrangement of horizontal connections
in tree shrew striate cortex. Journal of Neuroscience, 17, 2112 –
2127.
Brefczynski, J. A., & DeYoe, E. A. (1999). A physiological correlate
of the ‘spotlight’ of visual attention. Nature Neuroscience, 2,
370 – 374.
Cohn, T. E., & Lasley, D. J. (1974). Detectability of a luminance
increment: effect of spatial uncertainty. Journal of the Optical
Society of America, 64, 1715–1718.
Crist, R. E., Kapadia, M. K., Westheimer, G., & Gilbert, C. D.
(1997). Perceptual learning of spatial localization: specificity for
orientation, position and context. Journal of Neurophysiology, 78,
2889 – 2894.
Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective
attention. Annual Re6iew of Neuroscience, 18, 193–222.
Dresp, B. (1993). Bright lines and edges facilitate the detection of
small line targets. Spatial Vision, 7, 213–225.
Engel, F. L. (1971). Visual conspicuity, directed attention and retinal
locus. Vision Research, 11, 563–576.
Eriksen, C. W., & St. James, J. D. (1986). Visual attention within and
around the field of focal attention: a zoom lens model. Perception
and Psychophysics, 40, 225–240.
1225
Fahle, M., & Morgan, M. (1996). No transfer of perceptual learning
between similar stimuli in the same retinal position. Current
Biology, 6, 292 – 297.
Field, D. J., Hayes, A., & Hess, R. F. (1993). Contour integration by
the human visual system: evidence for a local ‘association field’.
Vision Research, 33, 173 – 193.
Gandhi, S. P., Heeger, D. J., & Boynton, G. M. (1999). Spatial
attention affects brain activity in human primary visual cortex.
Proceedings of the National Academy of Sciences, 96, 3314–3319.
Gilbert, C. D. (1992). Horizontal integration and cortical dynamics.
Neuron, 9, 1 – 13.
Gilbert, C. D., & Wiesel, T. N. (1979). Morphological and intracortical projections of functionally characterized neurons in the cat
visual cortex. Nature, 280, 120 – 125.
Gilbert, C. D., & Wiesel, T. N. (1983). Clustered intrinsic connections
in cat visual cortex. Journal of Neuroscience, 3, 1116 – 1133.
Gilbert, C. D., & Wiesel, T. N. (1989). Columnar specificity of
intrinsic horizontal and corticocortical connections in cat visual
cortex. Journal of Neuroscience, 9, 2432 – 2442.
Gilbert, C. D., & Wiesel, T. N. (1990). The influence of contextual
stimuli on the orientation selectivity of cells in primary visual
cortex. Vision Research, 30, 1689 – 1701.
Gilbert, C. D. (1994). Early perceptual learning. Proceedings of the
National Academy of Sciences, 91, 1195 – 1197.
Grossberg, S, & Mingolla, E. (1985). Neural dynamics of perceptual
grouping: textures, boundaries, and emergent segmentations. Perception and Psychophysics, 38, 141 – 171.
Haenny, P. E., & Schiller, P. H. (1988). State dependent activity in
monkey visual cortex. 1. Single cell activity in V1 and V4 on
visual tasks. Experimental Brain Research, 69, 225 – 244.
von der Heydt, R., & Peterhans, E. (1989). Mechanisms of contour
perception in monkey visual cortex. I. Lines of pattern discontinuity. Journal of Neuroscience, 9, 1731 – 1748.
Hirsch, J. A., & Gilbert, C. D. (1991). Synaptic physiology of
horizontal connections in the cat’s visual cortex. Journal of Neuroscience, 11, 1800 – 1809.
Hupe, J. M., James, A. C., Payne, B. R., Lomber, S. G., Girard, P.,
& Bullier, J. (1998). Cortical feedback improves discrimination
between figure and background by V1, V2 and V3 neurons.
Nature, 394, 784 – 787.
Ito, M., Westheimer, G., & Gilbert, C. D. (1998). Attention and
perceptual learning modulate contextual influences on visual perception. Neuron, 20, 1191 – 1197.
Ito, M., & Gilbert, C. D. (1999). Attention modulates contextual
influences in the primary visual cortex of alert monkeys. Neuron,
22, 593 – 604.
Kapadia, M. K., Ito, M., Gilbert, C. D., & Westheimer, G. (1995).
Improvement in visual sensitivity by changes in local context:
parallel studies in human observers and in V1 of alert monkeys.
Neuron, 15, 843 – 856.
Kapadia, M. K., Westheimer, G., & Gilbert, C. D. (1998). Spatial
distribution and dynamics of contextual interactions in cortical
area V1. Society for Neuroscience Abstracts, 24, 1980.
Kapadia, M. K., Westheimer, G., & Gilbert, C. D. (1999). Dynamics
of spatial summation in primary visual cortex of alert monkeys.
Proceedings of the National Academy of Sciences USA, 96, 12073–
12078.
Kastner, S., Nothdurft, H.-C., & Pigarev, I. N. (1997). Neural
correlates of pop-out in cat striate cortex. Vision Research, 37,
371 – 376.
Kisvarday, Z. F., Toth, E., Rausch, M., & Eysel, U. T. (1997).
Orientation specific relationship between populations of excitatory and inhibitory lateral connections in the visual cortex of the
cat. Cerebral Cortex, 7, 605 – 618.
Knierim, J. J., & Van Essen, D. C. (1992). Neuronal responses to
static texture patterns in are V1 of the alert macaque monkey.
Journal of Neuroscience, 67, 961 – 980.
1226
C. Gilbert et al. / Vision Research 40 (2000) 1217–1226
Kosslyn, S. M., Thompson, W. L, Kim, I. J., Rauch, S. L., & Alpert,
N. M. (1995). Topographical representations of mental images in
primary visual cortex. Nature, 378, 496–498.
Krose, B. J., & Julesz, B. (1989). The control and speed of shifts of
attention. Vision Research, 29, 1607–1619.
Kuffler, S. W. (1953). Discharge patterns and functional organization
of mammalian retina. Journal of Neurophysiology, 16, 37– 68.
LaBerge, D. (1983). Spatial extent of attention to letters and words.
Journal of Experimental Psychology: Human Perception & Performance, 9, 371 – 379.
Lamme, V. A. F. (1995). The neurophysiology of figure-ground
segregation in primary visual cortex. Journal of Neuroscience, 15,
1605 – 1615.
Levitt, J. B., & Lund, J. S. (1997). Contrast dependence of contextual
effects in primate visual cortex. Nature, 387, 73–76.
Li, C.-Y., & Li, W. (1994). Extensive integration field beyond the
classical receptive field of cat’s striate cortical neurons – Classification and tuning properties. Vision Research, 34, 2337–2355.
Lindblom, B., & Westheimer, G. (1992a). Uncertainty effects in
orientation discrimination of foveally seen lines in human observers. Journal of Physiology (London), 454, 1–8.
Lindblom, B., & Westheimer, G. (1992b). Spatial uncertainty in
stereoacuity tests: implications for clinical vision test design. Acta
Opthalmology, 70, 60–65.
Luck, S. J., Chelazzi, L., Hillyard, S. A., & Desimone, R. (1997).
Neural mechanisms of spatial attention in areas V1, V2 and V4 of
macaque visual cortex. Journal of Neurophysiology, 77, 24 – 42.
Maffei, L., & Fiorentini, A. (1976). The unresponsive regions of
visual cortical receptive fields. Vision Research, 16, 1131– 1139.
Malach, R., Amir, Y., Harel, M., & Grinvald, A. (1993). Relationship between intrinsic connections and functional architecture
revealed by optical imaging and in vivo targeted biocytin injections in primate striate cortex. Proceedings of the National
Academy of Sciences, 90, 10469–10473.
Martinez, A., Anllo-Vento, L, Sereno, M. I., Frank, L. R., Buxton,
R. B., Dubowitz, D. J., Wong, E. C., Hinrichs, H. J., & Hillyard,
S. A. (1999). Involvement of striate and extrastriate cortical areas
in spatial attention. Nature Neuroscience, 2, 364–369.
Motter, B. C. (1993). Focal attention produces spatially selective
processing in visual cortical areas V1, V2, and V4 in the presence
of competing stimuli. Journal of Neurophysiology, 70, 909 – 919.
Nakayama, K., & Joseph, J. S. (1997). Attention, pattern recognition,
and popout in visual search. In R. Parasuraman, The attenti6e
brain (pp. 279 – 298). Cambridge, MA: MIT Press.
Nelson, J. I., & Frost, B. J. (1978). Orientation-selective inhibition
from beyond the classic visual receptive field. Brain Research, 139,
359 – 365.
Nothdurft, H. C., & Li, C. Y. (1985). Texture discrimination: representation of orientation and luminance differences in cells of the
cat striate cortex. Vision Research, 25, 99–113.
Orban, G. A., Guylas, B., & Vogels, R. (1987). Influence of a moving
textural background on direction selectivity of cat striate neurons.
Journal of Neurophysiology, 57, 1767–1791.
Polat, U., & Sagi, D. (1993). Lateral interactions between spatial
channels: suppression and facilitation revealed by lateral masking
experiments. Vision Research, 33, 993–999.
.
Polat, U., & Sagi, D. (1994). The architecture of perceptual spatial
interactions. Vision Research, 28, 115 – 132.
Polat, U., Mizobi, K., Pettet, M. W., Kasamatsu, T., & Norcia, T.
(1998). Collinear stimuli regulate visual responses depending on
cell’s contrast threshold. Nature, 391, 580 – 584.
Rockland, K. S., & Lund, J. S. (1982). Widespread periodic intrinsic
connections in the tree shrew visual cortex. Science, 215, 1532–
1534.
Roelfesma, P. R., Lamme, V. A., & Spekreijse, H. (1998). Objectbased attention in the primary visual cortex of the macaque
monkey. Nature, 395, 376 – 381.
Rossi, A. F., Rittenhouse, C. D., & Paradiso, M. A. (1996). The
representation of brightness in primary visual cortex. Science, 273,
1104 – 1107.
Sagi, D., & Tanne, D. (1994). Perceptual learning: learning to see.
Current Opinion in Neurobiology, 4, 195 – 199.
Shiu, L.-P., & Pashler, H. (1992). Improvement in line orientation
discrimination is retinally local but dependent on cognitive set.
Perception & Psychophysics, 52, 582 – 588.
Sillito, A. M., Grieve, K. L., Jones, H. L., Cuderio, J., & Davis, J.
(1995). Visual cortical mechanisms detecting focal orientation
discontinuities. Nature, 378, 492 – 496.
Sireteanu, R., & Rettenbach, R. (1995). Perceptual learning in visual
search: fast, enduring, but non-specific. Vision Research, 35,
2037 – 2043.
Somers, D. C., Dale, A. M., Seiffert, A. E., & Tootell, R. B. H.
(1999). Functional MRI reveals spatially specific attentional modulation in human primary visual cortex. Proceedings of the National Academy of Sciences USA, 96, 1663 – 1668.
Tanaka, K., Hikosaka, K., Saito, H., Yukie, M., Fukada, Y., & Iwai,
E. (1986). Analysis of local and wide-field movements in the
superior temporal areas of the macaque monkey. Journal of
Neuroscience, 6, 134 – 144.
Treisman, A., Veira, A., & Hayes, A. (1992). Automaticity and
preattentive processing. American Journal of Psychology, 105,
341 – 362.
Ullman, S. (1990). Three-dimensional object recognition. Cold Spring
Harbor Symposium Quantitati6e Biology, 55, 889 – 898.
Vidyasagar, T. R. (1998). Gating of neuronal responses in macaque
primary visual cortex by an attentional spotlight. NeuroReport, 9,
1947 – 1952.
Weliky, M., Kandler, K., Fitzpatrick, D., & Katz, L. C. (1995).
Patterns of excitation and inhibition evoked by horizontal connections in visual cortex share a common relationship to orientation
columns. Neuron, 15, 541 – 552.
Wertheimer, M. (1923). Untersuchungen zur Lehre von der Gestalt
II, Psychologische Forshung, 4, 301 – 350 [translated as Laws of
organization in perceptual forms. In W. D. Ellis, A Source Book
of Gestalt Psychology (pp. 71 – 88). London, Harcourt, Brace and
Javanovich.
Wolfson, S. S., & Landy, M. S. (1999). Long range interactions
between oriented texture elements. Vision Research, 39, 933–945.
Wurtz, R. H., & Mohler, C. W. (1976). Enhancement of visual
responses in monkey striate cortex and frontal eye fields. Journal
of Neurophysiology, 39, 766 – 772.
Zipser, K., Lamme, V. A. F., & Schiller, P. H. (1996). Contextual
modulation in primary visual cortex. Journal of Neuroscience, 15,
7376 – 7389.